Visual Mapping Method for Autonomous Driving Sensor Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Simultaneous Localization and Mapping (SLAM) systems for autonomous driving are complex and time-consuming due to the lack of synchronization and physical alignment of sensors, making it difficult to generate accurate high-definition road maps in real-time.
Innovation Solution
A visual mapping method that maps feature points from camera images to lidar point cloud data, using a data generator to create and calibrate feature maps, allowing for accurate position estimation and synchronization of sensors like cameras and lidars.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional SLAM systems are used to generate high-definition road maps, then mapping accuracy can be achieved, but the system complexity and processing time increase significantly
Solution Approach 1:
The patent divides the mapping process into two distinct stages: offline feature extraction from pre-stored point cloud data, and online mapping by matching extracted features with current sensor data. This segmentation reduces the computational burden during real-time operation while maintaining high mapping accuracy through pre-computed feature relationships.
Solution Approach 2:
The patent performs feature extraction and relationship establishment in advance during an offline phase using pre-stored point cloud data. By preparing feature maps and spatial relationships beforehand, the system eliminates complex real-time computations during online operation, thereby reducing system complexity while preserving accuracy.
2Ease of manufacture
If sensors are not physically aligned and synchronized, then ease of installation is improved, but mapping precision deteriorates
Solution Approach 1:
The patent introduces a calibration process that acts as an intermediary between sensors with different positions and timing. By computing transformation parameters (rotation and translation matrices) that align sensor coordinate systems, the system reconciles data from misaligned sensors without requiring physical realignment, thus maintaining ease of installation while achieving precise mapping.
Solution Approach 2:
The patent compensates for sensor misalignment by dynamically adjusting parameter transformations rather than requiring physical position adjustments. Through calibration computations that determine rotation and translation parameters, the system mathematically realigns sensor data, achieving precise mapping while preserving the advantage of flexible sensor installation.
3Productivity
If real-time processing is required for autonomous driving, then productivity is improved, but measurement precision deteriorates due to computational limitations
Solution Approach 1:
The patent segments computational tasks into offline feature extraction and online feature matching. By pre-computing feature relationships from point cloud data, the system reduces real-time computational requirements while maintaining position estimation accuracy through efficient feature correspondence algorithms.
Solution Approach 2:
The patent creates a simplified representation (feature map) from the complex point cloud data by extracting and storing characteristic features. This copying approach allows the system to work with a reduced data representation during real-time operation, improving processing speed while preserving the essential information needed for accurate position estimation.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables efficient and accurate generation of feature maps, supporting real-time position estimation and reducing the complexity of SLAM systems, thereby enhancing the creation of high-definition road maps for autonomous driving.
Implementation Method 1
Light Detection and Ranging (LiDAR), and other sensors
Implementation Method 2
a camera that collects the shape and information of terrain features
Data Source
AI summary
Proposed is a visual mapping method for generating a feature map by mapping a feature point of an image captured by a camera to point cloud data acquired by a lidar. The method may include generating a first feature map based on point cloud data obtained from a lidar and an image captured from a camera, by a data generator, and generating a third feature map by mapping the first feature map on a second feature map generated through pre-stored point cloud data, by the data generator. The present method is a technology developed with support from the Ministry of Trade, Industry and Energy/Korea Planning and Evaluation Institute of Industrial Technology (Project No. 201792/Business name-Excellent enterprise research center promotion project (ATC+)/Project name-Development of real-time risk detection and mapping solution based on 3D scanning technology to ensure safety in autonomous driving).